harnessed 4.8.0 → 4.9.1

package/workflows/research/SKILL.zh-Hans.md

@@ -39,43 +39,13 @@ Sister `workflows/capabilities.yaml` 条目：
 
 ## 如何调用
 
-**优先路径**（master orchestrator）：按当前 stage 规定的顺序依次调用各子工作流的 slash command。每个子命令安装在 `<claude-home>/commands/<sub-name>.md`（由 `harnessed setup` 安装），自带双路径兜底。
+CC-native 编排。**不要** pipe 到 `harnessed run research` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-**兜底路径**（当上游未安装或无返回结果时）：使用 Task 工具 spawn 一个通用 subagent，prompt 如下：
+改用 `/research` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/research.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/research.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-> You are a **Research analyst**.
->
-> **Mission**: Multi-source investigation (docs / web search / codebase grep / library probe) producing a `findings.md` with citations, NOT speculation. Use `ctx7` for library docs, `tavily-mcp` / `exa-mcp` for web, `gh` CLI for GitHub artifacts, and codebase `Grep` for internal references.
->
-> **Default-suspect mode**: assume the change is broken / risky / incomplete until proven otherwise. Cite `file:line` for every finding; do not generalize.
->
-> **Review checklist**:
-> 1. Resolve each unknown claim to a citable source (URL, file:line, or `ctx7` doc id)
->
-> 2. Cite version explicitly when discussing library / framework APIs (training cutoff may be stale)
->
-> 3. Capture conflicting sources side-by-side; do not silently pick one
->
-> 4. Flag `OPEN: <question>` for items the user must decide; never paper over
->
-> 5. Persist results to `.planning/<phase>/findings.md` for cross-session handoff
->
-> **Output format**: structured report with severity-classified findings (verified / unverified / conflicting / open). One finding per line: `[severity] file:line — problem (one sentence); fix: suggested change`. If no findings, say so explicitly. No preamble, no end-of-report summary.
-
-（Role prompt 自包含 — 即使上游 `specialist` user-skill / plugin 未安装也可正常运行。）
-
-（`harnessed setup` 同时会在 `<claude-home>/commands/research.md` 安装一个 `research` Claude Code slash command，使 `/research` 成为真正的平台 slash command — 两个文件携带相同的双路径指令。之前 v3.4.x 中 `harnessed research --apply` CLI 的相关说明已移除；该子命令从未被实现。）
-
-## 如何调用
-
-使用 Bash 工具运行：
-
-```bash
-echo "$ARGUMENTS" | harnessed run research --task-stdin
-```
-
-如果 `$ARGUMENTS` 为空，运行 `harnessed run research`（不带 stdin pipe）。
-
-完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一个阶段。根据对话上下文决定是否调用 — 该提示仅供参考，不作强制指引。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->

package/workflows/retro/SKILL.md

@@ -49,17 +49,18 @@ Sister `workflows/capabilities.yaml` entries:
 
 ## How to invoke
 
-Use the Bash tool to run:
-
-```bash
-echo "$ARGUMENTS" | harnessed run retro --task-stdin
-```
-
-If `$ARGUMENTS` is empty, run `harnessed run retro` (no stdin pipe).
-
-After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
-
-<!-- harnessed-generated:v3.4.4 -->
+Orchestrated CC-natively. Do NOT pipe to `harnessed run retro` — that is the CI/headless
+path (an in-process SDK spawn that blocks the session, bypasses Agent Teams, and hangs when
+invoked from inside a Claude Code session).
+
+Run the `/retro` slash command instead (generated by `harnessed setup` at
+`~/.claude/commands/retro.md`). It drives the stage natively: `harnessed gates` → which
+subs fire, `harnessed prompt <sub>` → each spawn-ready prompt, then a CC-native subagent
+(Task / Agent tool) per fired sub, recording each outcome with `harnessed checkpoint`. The
+full state-machine steps live in `~/.claude/commands/retro.md`; if that file is absent,
+follow that same gates → prompt → spawn → checkpoint sequence yourself.
+
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/retro/SKILL.zh-Hans.md

@@ -49,17 +49,16 @@ Sister `workflows/capabilities.yaml` 条目：
 
 ## 如何调用
 
-使用 Bash 工具运行：
+CC-native 编排。**不要** pipe 到 `harnessed run retro` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-```bash
-echo "$ARGUMENTS" | harnessed run retro --task-stdin
-```
+改用 `/retro` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/retro.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/retro.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-如果 `$ARGUMENTS` 为空，运行 `harnessed run retro`（不带 stdin pipe）。
-
-完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一个阶段。根据对话上下文决定是否调用 — 该提示仅供参考，不作强制指引。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->
 
 ## 参考文档
 

package/workflows/task/auto/SKILL.md

@@ -59,17 +59,18 @@ Sister `workflows/capabilities.yaml`:
 
 ## How to invoke
 
-Use the Bash tool to run:
-
-```bash
-echo "$ARGUMENTS" | harnessed run task --task-stdin
-```
-
-If `$ARGUMENTS` is empty, run `harnessed run task` (no stdin pipe).
-
-After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
-
-<!-- harnessed-generated:v3.4.4 -->
+Orchestrated CC-natively. Do NOT pipe to `harnessed run task` — that is the CI/headless
+path (an in-process SDK spawn that blocks the session, bypasses Agent Teams, and hangs when
+invoked from inside a Claude Code session).
+
+Run the `/task` slash command instead (generated by `harnessed setup` at
+`~/.claude/commands/task.md`). It drives the stage natively: `harnessed gates` → which
+subs fire, `harnessed prompt <sub>` → each spawn-ready prompt, then a CC-native subagent
+(Task / Agent tool) per fired sub, recording each outcome with `harnessed checkpoint`. The
+full state-machine steps live in `~/.claude/commands/task.md`; if that file is absent,
+follow that same gates → prompt → spawn → checkpoint sequence yourself.
+
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/task/auto/SKILL.zh-Hans.md

@@ -57,19 +57,18 @@ Sister `workflows/capabilities.yaml`:
 
 - Slash command: `/task <text>` (bare per ADR 0030 namespace policy D-02 LOCK after `harnessed setup`)
 
-## How to invoke
+## 如何调用
 
-使用 Bash 工具运行：
+CC-native 编排。**不要** pipe 到 `harnessed run task` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-```bash
-echo "$ARGUMENTS" | harnessed run task --task-stdin
-```
+改用 `/task` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/task.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/task.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-若 `$ARGUMENTS` 为空，运行 `harnessed run task`（不带 stdin pipe）。
-
-执行完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一个阶段。根据对话上下文自行决定是否调用——该提示仅供参考，不具强制性。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/task/clarify/SKILL.md

@@ -56,17 +56,18 @@ phase-level conditional tool fire NOT 决定 phase 是否走)。
 
 ## How to invoke
 
-Use the Bash tool to run:
-
-```bash
-echo "$ARGUMENTS" | harnessed run task-clarify --task-stdin
-```
-
-If `$ARGUMENTS` is empty, run `harnessed run task-clarify` (no stdin pipe).
-
-After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
-
-<!-- harnessed-generated:v3.4.4 -->
+Orchestrated CC-natively. Do NOT pipe to `harnessed run task-clarify` — that is the CI/headless
+path (an in-process SDK spawn that blocks the session, bypasses Agent Teams, and hangs when
+invoked from inside a Claude Code session).
+
+Run the `/task-clarify` slash command instead (generated by `harnessed setup` at
+`~/.claude/commands/task-clarify.md`). It drives the stage natively: `harnessed gates` → which
+subs fire, `harnessed prompt <sub>` → each spawn-ready prompt, then a CC-native subagent
+(Task / Agent tool) per fired sub, recording each outcome with `harnessed checkpoint`. The
+full state-machine steps live in `~/.claude/commands/task-clarify.md`; if that file is absent,
+follow that same gates → prompt → spawn → checkpoint sequence yourself.
+
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/task/clarify/SKILL.zh-Hans.md

@@ -53,19 +53,18 @@ Phase 01-brainstorm 在 `phase.spec_ambiguous == true` 时条件性 fire `grill-
 无条件 fire（D-05 invokes_tools 与 OnClause 并存，但作用面不同——invokes_tools
 属于 phase 级别的条件性工具触发，不决定 phase 是否执行）。
 
-## How to invoke
+## 如何调用
 
-使用 Bash 工具运行：
+CC-native 编排。**不要** pipe 到 `harnessed run task-clarify` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-```bash
-echo "$ARGUMENTS" | harnessed run task-clarify --task-stdin
-```
+改用 `/task-clarify` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/task-clarify.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/task-clarify.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-若 `$ARGUMENTS` 为空，运行 `harnessed run task-clarify`（不带 stdin pipe）。
-
-执行完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一个阶段。根据对话上下文自行决定是否调用——该提示仅供参考，不具强制性。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/task/code/SKILL.md

@@ -62,17 +62,18 @@ marketplace).
 
 ## How to invoke
 
-Use the Bash tool to run:
-
-```bash
-echo "$ARGUMENTS" | harnessed run task-code --task-stdin
-```
-
-If `$ARGUMENTS` is empty, run `harnessed run task-code` (no stdin pipe).
-
-After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
-
-<!-- harnessed-generated:v3.4.4 -->
+Orchestrated CC-natively. Do NOT pipe to `harnessed run task-code` — that is the CI/headless
+path (an in-process SDK spawn that blocks the session, bypasses Agent Teams, and hangs when
+invoked from inside a Claude Code session).
+
+Run the `/task-code` slash command instead (generated by `harnessed setup` at
+`~/.claude/commands/task-code.md`). It drives the stage natively: `harnessed gates` → which
+subs fire, `harnessed prompt <sub>` → each spawn-ready prompt, then a CC-native subagent
+(Task / Agent tool) per fired sub, recording each outcome with `harnessed checkpoint`. The
+full state-machine steps live in `~/.claude/commands/task-code.md`; if that file is absent,
+follow that same gates → prompt → spawn → checkpoint sequence yourself.
+
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/task/code/SKILL.zh-Hans.md

@@ -59,19 +59,18 @@ Phase 01-code 根据 phase fact context 条件性 fire 3 个 mattpocock 招式
 遵循捆绑的「跨 session 恢复」模式 + R20.6 Manus-style 持久化。需要安装
 `planning-with-files` Claude Code plugin（通过 Claude Code plugin marketplace 安装）。
 
-## How to invoke
+## 如何调用
 
-使用 Bash 工具运行：
+CC-native 编排。**不要** pipe 到 `harnessed run task-code` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-```bash
-echo "$ARGUMENTS" | harnessed run task-code --task-stdin
-```
+改用 `/task-code` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/task-code.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/task-code.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-若 `$ARGUMENTS` 为空，运行 `harnessed run task-code`（不带 stdin pipe）。
-
-执行完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一个阶段。根据对话上下文自行决定是否调用——该提示仅供参考，不具强制性。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/task/deliver/SKILL.md

@@ -84,17 +84,18 @@ Claude Code plugin marketplace).
 
 ## How to invoke
 
-Use the Bash tool to run:
-
-```bash
-echo "$ARGUMENTS" | harnessed run task-deliver --task-stdin
-```
-
-If `$ARGUMENTS` is empty, run `harnessed run task-deliver` (no stdin pipe).
-
-After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
-
-<!-- harnessed-generated:v3.4.4 -->
+Orchestrated CC-natively. Do NOT pipe to `harnessed run task-deliver` — that is the CI/headless
+path (an in-process SDK spawn that blocks the session, bypasses Agent Teams, and hangs when
+invoked from inside a Claude Code session).
+
+Run the `/task-deliver` slash command instead (generated by `harnessed setup` at
+`~/.claude/commands/task-deliver.md`). It drives the stage natively: `harnessed gates` → which
+subs fire, `harnessed prompt <sub>` → each spawn-ready prompt, then a CC-native subagent
+(Task / Agent tool) per fired sub, recording each outcome with `harnessed checkpoint`. The
+full state-machine steps live in `~/.claude/commands/task-deliver.md`; if that file is absent,
+follow that same gates → prompt → spawn → checkpoint sequence yourself.
+
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/task/deliver/SKILL.zh-Hans.md

@@ -82,19 +82,18 @@ subtask 标记为完成——对应 Phase 01-code progress update 模式，是 S
 的最后一次调用。需要安装 `planning-with-files` Claude Code plugin（通过
 Claude Code plugin marketplace 安装）。
 
-## How to invoke
+## 如何调用
 
-使用 Bash 工具运行：
+CC-native 编排。**不要** pipe 到 `harnessed run task-deliver` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-```bash
-echo "$ARGUMENTS" | harnessed run task-deliver --task-stdin
-```
+改用 `/task-deliver` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/task-deliver.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/task-deliver.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-若 `$ARGUMENTS` 为空，运行 `harnessed run task-deliver`（不带 stdin pipe）。
-
-执行完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一个阶段。根据对话上下文自行决定是否调用——该提示仅供参考，不具强制性。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/task/test/SKILL.md

@@ -65,17 +65,18 @@ fix → regression-test), 测试通过则 skip diagnose entirely。
 
 ## How to invoke
 
-Use the Bash tool to run:
-
-```bash
-echo "$ARGUMENTS" | harnessed run task-test --task-stdin
-```
-
-If `$ARGUMENTS` is empty, run `harnessed run task-test` (no stdin pipe).
-
-After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
-
-<!-- harnessed-generated:v3.4.4 -->
+Orchestrated CC-natively. Do NOT pipe to `harnessed run task-test` — that is the CI/headless
+path (an in-process SDK spawn that blocks the session, bypasses Agent Teams, and hangs when
+invoked from inside a Claude Code session).
+
+Run the `/task-test` slash command instead (generated by `harnessed setup` at
+`~/.claude/commands/task-test.md`). It drives the stage natively: `harnessed gates` → which
+subs fire, `harnessed prompt <sub>` → each spawn-ready prompt, then a CC-native subagent
+(Task / Agent tool) per fired sub, recording each outcome with `harnessed checkpoint`. The
+full state-machine steps live in `~/.claude/commands/task-test.md`; if that file is absent,
+follow that same gates → prompt → spawn → checkpoint sequence yourself.
+
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/task/test/SKILL.zh-Hans.md

@@ -63,19 +63,18 @@ SDK，用户发出明确信号时可切换至 mattpocock /tdd 别名路径。
 测试失败时进入 diagnose 循环（reproduce → minimise → hypothesise → instrument →
 fix → regression-test），测试通过则完全跳过 diagnose。
 
-## 调用方式
+## 如何调用
 
-使用 Bash 工具运行：
+CC-native 编排。**不要** pipe 到 `harnessed run task-test` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-```bash
-echo "$ARGUMENTS" | harnessed run task-test --task-stdin
-```
+改用 `/task-test` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/task-test.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/task-test.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-若 `$ARGUMENTS` 为空，运行 `harnessed run task-test`（不带 stdin pipe）。
-
-执行完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一阶段操作。是否调用取决于对话上下文——该提示仅供参考，不具强制性。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->
 
 ## 参考资料
 

package/workflows/verify/auto/SKILL.md

@@ -66,17 +66,18 @@ Sister `workflows/capabilities.yaml`:
 
 ## How to invoke
 
-Use the Bash tool to run:
-
-```bash
-echo "$ARGUMENTS" | harnessed run verify --task-stdin
-```
-
-If `$ARGUMENTS` is empty, run `harnessed run verify` (no stdin pipe).
-
-After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
-
-<!-- harnessed-generated:v3.4.4 -->
+Orchestrated CC-natively. Do NOT pipe to `harnessed run verify` — that is the CI/headless
+path (an in-process SDK spawn that blocks the session, bypasses Agent Teams, and hangs when
+invoked from inside a Claude Code session).
+
+Run the `/verify` slash command instead (generated by `harnessed setup` at
+`~/.claude/commands/verify.md`). It drives the stage natively: `harnessed gates` → which
+subs fire, `harnessed prompt <sub>` → each spawn-ready prompt, then a CC-native subagent
+(Task / Agent tool) per fired sub, recording each outcome with `harnessed checkpoint`. The
+full state-machine steps live in `~/.claude/commands/verify.md`; if that file is absent,
+follow that same gates → prompt → spawn → checkpoint sequence yourself.
+
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/verify/auto/SKILL.zh-Hans.md

@@ -66,17 +66,16 @@ Sister `workflows/capabilities.yaml`：
 
 ## 如何调用
 
-使用 Bash 工具运行：
+CC-native 编排。**不要** pipe 到 `harnessed run verify` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-```bash
-echo "$ARGUMENTS" | harnessed run verify --task-stdin
-```
+改用 `/verify` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/verify.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/verify.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-若 `$ARGUMENTS` 为空，运行 `harnessed run verify`（不带 stdin pipe）。
-
-执行完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一阶段操作。是否调用取决于对话上下文——该提示仅供参考，不具强制性。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->
 
 ## 参考资料
 

package/workflows/verify/code-review/SKILL.md

@@ -50,17 +50,18 @@ Sister `workflows/judgments/parallelism-gate.yaml`:
 
 ## How to invoke
 
-Use the Bash tool to run:
-
-```bash
-echo "$ARGUMENTS" | harnessed run verify-code-review --task-stdin
-```
-
-If `$ARGUMENTS` is empty, run `harnessed run verify-code-review` (no stdin pipe).
-
-After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
-
-<!-- harnessed-generated:v3.4.4 -->
+Orchestrated CC-natively. Do NOT pipe to `harnessed run verify-code-review` — that is the CI/headless
+path (an in-process SDK spawn that blocks the session, bypasses Agent Teams, and hangs when
+invoked from inside a Claude Code session).
+
+Run the `/verify-code-review` slash command instead (generated by `harnessed setup` at
+`~/.claude/commands/verify-code-review.md`). It drives the stage natively: `harnessed gates` → which
+subs fire, `harnessed prompt <sub>` → each spawn-ready prompt, then a CC-native subagent
+(Task / Agent tool) per fired sub, recording each outcome with `harnessed checkpoint`. The
+full state-machine steps live in `~/.claude/commands/verify-code-review.md`; if that file is absent,
+follow that same gates → prompt → spawn → checkpoint sequence yourself.
+
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/verify/code-review/SKILL.zh-Hans.md

@@ -50,17 +50,16 @@ Sister `workflows/judgments/parallelism-gate.yaml`：
 
 ## 如何调用
 
-使用 Bash 工具运行：
+CC-native 编排。**不要** pipe 到 `harnessed run verify-code-review` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-```bash
-echo "$ARGUMENTS" | harnessed run verify-code-review --task-stdin
-```
+改用 `/verify-code-review` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/verify-code-review.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/verify-code-review.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-若 `$ARGUMENTS` 为空，运行 `harnessed run verify-code-review`（不带 stdin pipe）。
-
-执行完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一阶段操作。是否调用取决于对话上下文——该提示仅供参考，不具强制性。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->
 
 ## 参考资料
 

package/workflows/verify/design/SKILL.md

@@ -52,17 +52,18 @@ Sister `workflows/judgments/stage-routing.yaml`:
 
 ## How to invoke
 
-Use the Bash tool to run:
-
-```bash
-echo "$ARGUMENTS" | harnessed run verify-design --task-stdin
-```
-
-If `$ARGUMENTS` is empty, run `harnessed run verify-design` (no stdin pipe).
-
-After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
-
-<!-- harnessed-generated:v3.4.4 -->
+Orchestrated CC-natively. Do NOT pipe to `harnessed run verify-design` — that is the CI/headless
+path (an in-process SDK spawn that blocks the session, bypasses Agent Teams, and hangs when
+invoked from inside a Claude Code session).
+
+Run the `/verify-design` slash command instead (generated by `harnessed setup` at
+`~/.claude/commands/verify-design.md`). It drives the stage natively: `harnessed gates` → which
+subs fire, `harnessed prompt <sub>` → each spawn-ready prompt, then a CC-native subagent
+(Task / Agent tool) per fired sub, recording each outcome with `harnessed checkpoint`. The
+full state-machine steps live in `~/.claude/commands/verify-design.md`; if that file is absent,
+follow that same gates → prompt → spawn → checkpoint sequence yourself.
+
+<!-- harnessed-generated:v4.9.1 -->
 
 ## References
 

package/workflows/verify/design/SKILL.zh-Hans.md

@@ -50,19 +50,18 @@ Sister `workflows/judgments/stage-routing.yaml`：
 - 创意补充 / 不要 AI 味 → `frontend-design`
 - 用户明示「独特 / 不要 AI 感」→ frontend-design 主导，否则 ui-ux-pro-max 优先
 
-## 调用方式
+## 如何调用
 
-使用 Bash 工具运行：
+CC-native 编排。**不要** pipe 到 `harnessed run verify-design` —— 那是 CI/headless 路径(in-process
+SDK spawn,会阻塞 session、绕过 Agent Teams,在 Claude Code 内部调用时还会挂死)。
 
-```bash
-echo "$ARGUMENTS" | harnessed run verify-design --task-stdin
-```
+改用 `/verify-design` slash command(由 `harnessed setup` 生成于 `~/.claude/commands/verify-design.md`)。
+它以 CC-native 方式驱动:`harnessed gates` 决定哪些 sub fire,`harnessed prompt <sub>` 给出每个
+spawn-ready prompt,然后用 CC-native subagent(Task / Agent 工具)逐个 spawn 已 fire 的 sub,
+每个结果用 `harnessed checkpoint` 记录。完整 state-machine 步骤见 `~/.claude/commands/verify-design.md`;
+若该文件不存在,自行按 gates → prompt → spawn → checkpoint 同序执行。
 
-若 `$ARGUMENTS` 为空，运行 `harnessed run verify-design`（不带 stdin pipe）。
-
-执行完成后，Bash 输出会在 stderr 打印 `Next:` 提示，建议下一阶段操作。是否继续调用，请根据对话上下文自行判断——该提示仅供参考，并非强制指令。
-
-<!-- harnessed-generated:v3.4.4 -->
+<!-- harnessed-generated:v4.9.1 -->
 
 ## 参考资料